2021.03.12.435067.Full.Pdf

Home , Canavanine

bioRxiv preprint doi: https://doi.org/10.1101/2021.03.12.435067; this version posted March 12, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1 A Fragment-based approach to assess the ligandability of ArgB, ArgC, ArgD and ArgF in the

2 L-arginine biosynthetic pathway of Mycobacterium tuberculosis

4 Pooja Gupta1, Sherine E. Thomas1, James Cory-Wright1, Víctor Sebastián-Pérez1,2, Ailidh Burgess1,

5 Emma Cattermole1, Clio Meghir1, Chris Abell3, Anthony G. Coyne3, William R. Jacobs Jr. 4 Tom L.

6 Blundell1, Sangeeta Tiwari5* and Vítor Mendes1#*.

8 1 Department of Biochemistry, University of Cambridge, 80 Tennis Court Road, Cambridge, CB2 1GA,

9 UK

10 2 Centro de Investigaciones Biológicas Margarita Salas (CSIC), Ramiro de Maeztu 9, 28040 Madrid,

11 Spain.

12 2 Department of Microbiology and Immunology, Albert Einstein College of Medicine, Bronx, NY 10461,

13 USA

14 3 Yusuf Hamied Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge, CB2

15 1EW, UK

16 4 Department of Microbiology and Immunology, Albert Einstein College of Medicine, Bronx, NY 10461,

17 USA.

18 5 Department of Biological Sciences & Border Biomedical Research Centre, University of Texas at El

19 Paso, El Paso, TX 79968, USA.

20 # Current affiliation: MRC-Laboratory of Molecular Biology, Molecular Immunity Unit, Francis Crick

21 Ave, Cambridge, CB2 0QH, UK.

23 * To whom correspondence should be addressed

24 Vitor Mendes: [email protected]; +44 1223267723

25 Sangeeta Tiwari: [email protected]; +1 9157476889

27 Abstract

28 The L-arginine biosynthesis pathway consists of eight enzymes that catalyse the conversion of L-

29 glutamate to L-arginine, appears to be attractive target for anti-Tuberculosis (TB) drug discovery.

30 Starvation of M. tuberculosis deleted for either argB or argF genes led to rapid sterilization of these

31 strains in mice while Chemical inhibition of ArgJ with Pranlukast was also found to clear chronic M.

32 tuberculosis infection in animal models. In this work, the ligandability of four enzymes of the pathway

33 ArgB, ArgC, ArgD and ArgF is explored using a fragment-based approach. We reveal several hits for

34 these enzymes validated with biochemical and biophysical assays, and X-ray crystallographic data,

35 which in the case of ArgB were further confirmed to have on-target activity against M. tuberculosis.

36 These results demonstrate the potential of more enzymes in this pathway to be targeted with

37 dedicated drug discovery programmes.

39 Keywords

40 ArgB, ArgC, ArgD, ArgF, Mycobacterium tuberculosis, FBDD

42 Abbreviations

43 TB – tuberculosis; FBDD – Fragment-based drug discovery; DSF – Differential scanning fluorimetry;

44 ASU - asymmetric unit; SPR – Surface plasmon resonance; NMR – Nuclear magnetic resonance; ITC –

45 Isothermal titration calorimetry.

53 1. Introduction

54 Despite the availability of effective chemotherapy, tuberculosis (TB) remains a leading infectious cause

55 of morbidity and mortality worldwide. In 2019, an estimated 1.2 million deaths were caused by TB,

56 and an additional 208,000 were a result of HIV-TB co-infection (1). Simultaneously, the existing

57 multidrug treatment regimen has a success rate of 85% in drug-sensitive TB cases (in the 2018 cohort),

58 drug toxicity, a long treatment duration, and resulting patient non-compliance, as well as

59 incompatibility with antiretroviral therapy all compromise its effectiveness. Alarmingly, the

60 emergence of multi-drug resistant (MDR) and extensively-drug-resistant (XDR) strains of

61 Mycobacterium tuberculosis has further undermined the efficacy of current antitubercular therapy:

62 only 57% of MDR cases were successfully treated worldwide in the 2017 cohort. New antitubercular

63 agents are therefore urgently required and novel chemical scaffolds and mechanisms of action must

64 be identified that can shorten therapy and circumvent development of drug resistance. While many

65 drugs can be bacteriocidal, M. tuberculosis has the ability to generate subpopulations that enter into

66 a persister state making them phenotypically drug resistant (2). The consideration of preventing

67 persister formation or killing persisters needs to be addressed in future drug discovery campaigns

68 against M .tuberculosis.

69 M. tuberculosis, like the leprosy bacillus, has retained its ability to make all 20 amino acids and most

70 vitamins. This retention of these biosynthetic genes reflect a evolutionary pressure suggesting the

71 pathogenic mycobacteria have chosen not to obtain amino acids or many vitamins from the host and

72 has thus been described as an autarkic lifestyle (3). However, not all amino acid auxotrophies behave

73 the same. Several amino acid auxotrophs were found to have attenuated virulence inside host

74 organisms, suggesting that while enzymes in amino acid biosynthetic pathways are essential in vitro,

75 the pathogen can scavenge amino acids (albeit insufficiently) inside the host and survive (4-9).

76 However, it was shown that methionine and arginine auxotrophs of M. tuberculosis are rapidly

77 sterilised in both immunocompetent and immunodeficient (SCID) mice without the appearance of

78 suppressor/bypass mutants (3, 10). Despite the presence of two arginine transporters in M.

79 tuberculosis (11, 12) and sufficiently high serum concentrations of arginine in the host (13), the

80 virulence of ΔargB or ΔargF mutants is entirely abolished as arginine deprivation results in extensive

81 oxidative damage (10). The case for drug discovery approaches to target arginine biosynthetic

82 enzymes is further bolstered by work demonstrating that chemical inhibition of ArgJ with Pranlukast,

83 a cysteinyl leukotriene receptor-1 antagonist use to treat asthmatic exacerbations, cleared a chronic

84 M. tuberculosis infection in BALB/c mice (14). The arginine biosynthesis pathway consists of eight

85 different enzymes (Figure 1A) all considered to be essential for M. tuberculosis growth in vitro (15).

86 Except argA which encodes the first enzyme of the pathway, all other genes are present in a single

87 operon that also includes the repressor argR (Figure 1B).

88 Fragment based drug discovery (FBDD) is now an established lead-generation strategy in both industry

89 and academia, having yielded over 30 compounds in clinical trials, including approved cancer drugs

90 like vemurafenib, Kisqali, Balversa and venetoclax (16). This approach consists of screening a library

91 of small molecules (150-300 Da) against a target of interest using biophysical, biochemical and

92 structural biology methods. The low complexity of fragments allows for efficient exploration of the

93 chemical space of the target, often revealing unexpected binding sites in proteins. Although fragments

94 often bind weakly, they tend to bind to hotspot regions of the target, forming well defined interactions

95 that allow subsequent elaboration into larger, drug-like molecules (17, 18). Our group and a few others

96 have pioneered using this approach against different mycobacterial species and different protein

97 targets with varying degrees of success (18-26).

98 Using this approach, we have screened four enzymes of the arginine biosynthesis pathway not yet

99 explored drug discovery programmes: ArgB, ArgC, ArgD and ArgF. Herein we report the structures of

100 the four enzymes in complex with fragments hits, including a novel allosteric site of ArgB and allosteric

101 inhibitors of this enzyme. Importantly, this work also assesses the potential of these enzymes as

102 candidates of future drug discovery programmes.

103

104 2. Materials and Methods

105 2.1. Molecular Cloning

106 The argB gene was amplified from chromosomal DNA of M. tuberculosis H37Rv strain obtained from

107 ATCC (ATCC25618D-2) while the ORFs of argC, argD and argF were purchased as E. coli codon-

108 optimised synthetic gene strings through the ThermoFisher GeneArt Gene Synthesis service. The argB

109 gene was cloned into pHAT4 (27) using NcoI and XhoI sites. The gene strings of argC, argD and argF

110 were cloned into a pET28a vector (modified to include an N-terminal 6xhis SUMO) (28) using BamHI

111 and HindIII restriction sites. All constructs were confirmed by sequencing.

112

113 2.2. Protein expression and purification

114 250 mL of autoclaved 2xYT broth (Formedium) prepared in distilled water, containing 100 μg/mL

115 ampicillin for pHAT4:argB or 30 μg/mL kanamycin for pET28a:argC/argD/argF, was inoculated with E.

116 coli BL21(DE3) containing the respective expression construct, and incubated at 37 °C with 220 rpm

117 shaking overnight. This primary culture was used the following day to inoculate 6 flasks containing 1 L

118 2xYT broth and the appropriate antibiotic, and the inoculated flasks were incubated under similar

119 conditions until the OD600nm reached 0.8-1. Overexpression was induced by the addition of 0.5 mM

120 isopropyl β-D-1-thiogalactopyranoside (IPTG). Thereafter, the flasks were incubated at 20 °C with 220

121 rpm shaking overnight.

122 Cells were harvested by centrifugation at 4200 rpm, 4 °C for 20 minutes in a Beckman Coulter

123 ultracentrifuge. The cell pellets were re-suspended in 50 mL of Buffer A (Table S1), also containing 1

124 tablet of cOmplete™, Mini, EDTA-free Protease Inhibitor Cocktail (Roche, Merck), DNase I (Sigma-

125 Aldrich) and 5 mM MgCl2. The cells were lysed by ultrasonication for ~ 8 minutes (pulse on for 20 secs,

126 pulse off for 30 secs, 55% amplitude), the suspensions were kept in an ice bath throughout. The cell

127 lysates were clarified by centrifugation (27000 g, 4 °C for 40 minutes), and the supernatants were

128 syringe-filtered (0.45 μm membrane) to remove any cell debris.

129 The filtered lysates were subjected to IMAC using a His-Trap 5 mL Nickel column (GE Healthcare Life

130 Sciences) on an ÄKTA Pure system (GE Healthcare Life Sciences), equilibrated with buffer A (Table S1).

131 Isocratic elution was performed using buffer B (Table S1). Proteins were dialysed in buffer C at 4 °C

132 and tags were cleaved overnight by adding TEV protease for ArgB or Ulp1 protease, for ArgC, ArgD

133 and ArgF, both at 1:100 ratio.

134 The dialysed proteins were concentrated to a <5 mL volume using a 30 kDa MWCO Vivaspin 20

135 centrifugal concentrator (Sartorius) at 5000 g, 4 °C and injected onto a HiLoad 16/600 Superdex 200

136 gel filtration column (GE Life Sciences) equilibrated with buffer C (Table S1). Elution fractions

137 corresponding to the peak of interest in the chromatogram were pooled and fraction purity was

138 assessed by SDS-PAGE. The purest fractions of ArgB and ArgF were pooled and concentrated to 20

139 mg.ml-1. Pooled fractions of ArgC were further dialysed into the final storage buffer (5 mM Tris-HCl pH

140 7.4, 50 mM NaCl) overnight at 4 °C, rescued the next day and concentrated to 6.5 mg/mL. ArgD

141 fractions were pooled and aqueous pyridoxal-5’-phosphate (PLP, Sigma-Aldrich) was added (2 mM

142 final PLP concentration). Overnight dialysis into the storage buffer (50 mM Tris-HCl pH 7, 100 mM

143 NaCl) was carried out at 4 °C to remove excess PLP. The PLP-saturated protein was rescued the next

144 day and concentrated to 14 mg/mL. All proteins were flash frozen in liquid N2 and stored at -80 °C.

145

146

147

148 2.3. Differential Scanning Fluorimetry

149 Fragment screening was carried out in a 96-well PCR plate using a CFX Connect Real-time PCR

150 Detection System (Bio-Rad) for DSF. For ArgB, each 25 μL reaction mixture contained 10 µM ArgB, 100

151 mM HEPES (pH 7.5), 200 mM NaCl, 5x SYPRO Orange, 5% DMSO (v/v), and fragments at 5 mM. For

152 ArgF each 25 μL reaction mixture contained 5 µM ArgF, 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 5x

153 SYPRO Orange dye, 5 mM fragments and 5% DMSO (v/v). For ArgC and ArgD, the 25 μL reaction

154 volume consisted of the following: 2.5 μM ArgC/5 μM ArgD, 100 mM sodium phosphate pH 7, 200

155 mM NaCl, 5x SYPRO Orange, and 5 mM fragments (960 fragment library). The protocol implemented

156 increased temperature by 0.5 °C after every 30 seconds, going from 25 °C to 95 °C and measuring

157 SYPRO Orange fluorescence for each temperature cycle. The melt curve RFU (relative fluorescence

158 units) and derivative -d(RFU)/dT values were analysed and plotted using a macros-enabled Excel

159 Workbook: the minima of the melt curves were recorded as the melting temperature (Tm) of the

160 enzymes in the presence of each fragment. The Tm of the reference control (protein in the presence

161 of DMSO) was subtracted from all the readings to calculate ΔTm.

162

163 2.4. Surface plasmon resonance

164 Low molecular weight (LMW) screening with the DSF fragment hits was carried out using the T200

165 Biacore instrument (GE Healthcare Life Sciences). A series S CM5 sensor chip (GE Healthcare Life

166 Sciences) was used for the immobilisation of ArgC on the carboxymethylated dextran matrix through

167 amine coupling. A 25 μg/mL ArgC dilution was prepared in the optimal coupling buffer (Sodium acetate

168 pH 5), and immobilisation was performed by manual instructions. The immobilised ArgC was tested

169 using dilution series (19 μM to 2.5 mM) of NADP+ and NADPH. 50 mM fragment DMSO stocks were

170 used to prepare 1 mM dilutions in the SPR buffer consisting of 10 mM sodium phosphate pH 7.0 , 150

171 mM NaCl and 2% (v/v) DMSO. Each NADP+/NADPH and fragment dilution was injected once at a flow

172 rate of 30 μL/min for a contact time of 30 seconds, SPR Running buffer, consisting of 10 mM sodium

173 phosphate pH 7.0 and 150 mM NaCl, was passed for 320 seconds at the same flow rate, and 50%

174 DMSO (diluted in SPR running buffer) was injected at the end of the cycle to remove undissociated

175 analyte. Solvent correction was carried out to account for DMSO mismatch between the analyte

176 dilutions and the SPR running buffer.

177 A 30 μg/mL ArgD dilution was prepared in sodium acetate pH 4 buffer for immobilisation. Following

178 the ethanolamine neutralisation step, 1 mM PLP (prepared in the SPR buffer) was injected for a

179 contact time of 420 seconds to ensure saturation of PLP-binding sites. The immobilised ArgD was

180 tested using dilution series (19 μM to 2.5 mM) of L-glutamate, N-acetylornithine and L-ornithine.

181 Screening was also carried out against PLP-unsaturated ArgD.

182

183 2.5. Ligand-observed NMR

184 All NMR experiments were carried out at 298 K using a Bruker Avance 600 MHz spectrometer with a

185 Triple Resonance Inverse (TCI) Automatic Tuning and Matching (ATM) cryoprobe. T2 relaxation-

186 filtered one-dimensional NMR spectroscopy experiments incorporated a CPMG67 spin-lock time of

187 200 ms before the acquisition period. Samples (600 µL) containing 2 mM fragment in the absence and

188 presence of 10 µM ArgB were prepared in buffer containing 20 mM potassium phosphate at pH 7.4

189 and 50 mM NaCl. Additionally, 2% v/v d6-DMSO was present in all samples for fragment solubilisation

190 and field-frequency locking. Displacement experiments were carried out in the same manner by

191 adding 1 mM each of ATP, and N-acetyl-L-glutamic acid or L-arginine to the samples containing 2 mM

192 fragment and 10 µM ArgB. The samples were loaded into 5 mm NMR tubes (Wilmad, 526-PP) for

193 measurement, and the resulting spectra were analysed using TopSpin v. 3.5 (Bruker).

194

195 2.6. Crystallisation of the apoenzymes

196 Crystallisation screening and optimisation for all the enzymes was performed at 18 °C with the sitting

197 drop vapour diffusion method using a Mosquito robot (TTP-Labtech) to setup the crystallisation

198 experiments. For apo ArgB, 300 nL of pure protein at 10 mg.ml-1 was mixed with an equal volume of

199 reservoir solution and equilibrated against 85 µl of the reservoir solution. The selected condition was

200 obtained in Wizards Classic 1&2 crystallisation screen (Rigaku), well G5 (1260 mM ammonium

201 sulphate, 100 mM CHES pH 9.5 and 200 mM NaCl). Crystals appeared after 2 days in this condition

202 and diffracted up to 1.8 Å resolution. For ArgF, 200 nL of pure protein at 20 mg.mL-1 was mixed with

203 an equal volume of reservoir solution. An initial crystallization condition was identified I the Wizards

204 Classic 3&4 crystallisation screen (Rigaku), well F2 (40% PEG400, 100 mM Tris-HCl pH 7.5 and 200 mM

205 Lithium sulphate). However, crystals obtained in this condition contained two lattices with different

206 orientations and no structural solution could be found despite good quality diffraction. These crystals

207 were ground to produce seeds and a new crystallisation screening was performed using 200 nL of ArgF

208 at 20 mg.mL-1 mixed with equal volume of reservoir solution and 50 nL of seed solution. After several

209 rounds of optimisation, the final crystallisation condition consisted of 150 mM ammonium dihydrogen

210 phosphate and 10 mM praseodymium acetate. Crystals appeared after a 2 days and diffracted up to

211 1.8 Å resolution.

212 A previously reported crystallisation condition for ArgC (29) was reproduced with some modifications,

213 but the crystals were found to not be suitable for fragment soaking experiments. A new crystallisation

214 screen was therefore performed using 200 nL of pure ArgC at 6.5 mg.ml-1 mixed with an equal volume

215 of reservoir solution. Well A8 from the BCS screen (Molecular Dimensions) produced crystals

216 diffracting to 1.54 Å. This condition consisted of 0.1 M phosphate/citrate buffer pH 5.5 and 20% PEG

217 Smear High (PEG 6K, 8K, 10K) and was optimised to remove the cryoprotection step by adding 20%

218 glycerol. A second condition with a neutral pH more amenable to soaking based on the previous

219 condition was also optimised and consisted of 0.1 M Bis-Tris pH 7, 17% PEG Smear High, 70 mM

220 phosphate/citrate pH 5.6, 20% glycerol.

221 For ArgD, 200 nL PLP-saturated enzyme 14 mg.ml-1 was mixed with an equal volume of reservoir

222 solution. A condition was found in PEG Smear BCS screen (Molecular Dimensions) well F6 (0.1 M Bis-

223 Tris Propane pH 8.5, 18% PEG Smear High (PEGs 6K, 8K, 10K), 0.2 M Ammonium nitrate). The final

224 optimized conditions consisted of 0.1 M Bis-Tris Propane pH 8.5, 18% PEG Smear High, 0.2 M

225 ammonium nitrate and 10 mM nickel chloride (additive).

226

227 2.7. Crystal soaking and co-crystallisation with natural ligands and fragment hits

228 To obtain ligand-bound structures, soaking was performed in the crystallisation conditions described

229 above for each protein using the hanging drop vapour diffusion method.

230 For ArgB, 1.5 µL of protein storage buffer containing 20 mM of ligand was mixed with 1.5 µL of

231 reservoir solution, and drops were left to equilibrate against 500 µL of reservoir solution for 3 days.

232 Crystals were then transferred to the drops and incubated for 16 h. A cryogenic solution was prepared

233 by adding ethylene glycol up to 27.5% v/v to the mother liquor. Crystals were briefly transferred to

234 this solution, flash-frozen in liquid N2, and stored for data collection. To obtain an ArgB-NAG complex,

235 co-crystallization with 2 mM NAG was performed instead. Crystals for ArgB-NAG complex were

236 obtained in Wizard Classic 1&2 screen (Rigaku), solution B6, and were flash-frozen in liquid N2 after a

237 brief soak in a solution containing mother liquor and 27.5% ethylene glycol.

238 ArgC crystals grown in pH 5.5 were first soaked in 1.5 µL drops containing the mother liquor and 5 mM

239 NADP+ for 2 hours in hanging drops that were equilibrated against a reservoir of 500 µL. Thereafter,

240 the crystals were transferred to drops containing the crystallisation condition and an SPR-validated

241 fragments (20 mM, 10% DMSO), which were equilibrated against 500 µL of mother liquor also

242 containing a corresponding percentage of DMSO overnight at 19 °C. ArgC crystals grown in pH 7 were

243 soaked with 5 mM NADP+ only for 5-10 minutes due to the rapid development of cracks, and

244 transferred to the fragment soaking drops for 5-10 minutes from where they were fished and frozen.

245 ArgD crystals were soaked with fragments at a concentration of 50 mM overnight in otherwise the

246 same manner as ArgC crystals grown in pH 5.5. A cryogenic solution was prepared by adding 30%

247 ethylene glycol to the mother liquor. Crystals were briefly transferred to this solution and flash-frozen

248 in liquid N2.

249 ArgF crystals were soaked in drops containing crystallization condition and 20 mM of ligand and

250 equilibrated against 500 µL of reservoir solution for 16h. A cryogenic solution was prepared by adding

251 25% ethylene glycol to the mother liquor. Crystals were briefly transferred to this solution and flash-

252 frozen in liquid nitrogen.

253

254 2.8. X-ray data collection and processing

255 X-ray diffraction data (single-wavelength anomalous diffraction) were collected on beamlines i02, i03,

256 i04, i04-1 and i24 at the Diamond Light Source (DLS), UK and on id30B at The European Synchrotron

257 Radiation Facility (ESRF). Diffraction data were processed and reduced with autoPROC from Global

258 Phasing Limited (30) or Xia2 (31). The apo-form of ArgB was crystallized in the R32 spacegroup with

259 one protomer per asymmetric unit (ASU) and the ArgB:NAG complex in the P63 spacegroup with two

260 protomers per ASU. ArgF was crystallized in the P21 spacegroup with 6 protomers in the ASU. ArgC

261 was crystallized in C2 and P21 spacegroups, with 2 and 4/8 protomers per ASU respectively. ArgD was

262 crystallised in the P21 space group as well but with 4 protomers in the ASU.

263 Initial phases were determined with PHASER (32) from PHENIX software package (33) using the M.

264 tuberculosis ArgB structure (PDB: 2AP9), M. tuberculosis ArgF structure (PDB: 2P2G), M. tuberculosis

265 ArgC structure (PDB: 2I3G) and E. coli Succinyl-ornithine transaminase (AstC, 42% sequence identity,

266 PDB: 4ADB) as a search models, respectively for ArgB, ArgF, ArgC and ArgD. Model building was done

267 with Coot (34), and refinement was performed in PHENIX (33, 35) for ArgB, ArgF, and ArgC. For ArgD,

268 after the initial molecular replacement solution and a single cycle of refinement, PHENIX AutoBuild

269 was used to generate a model for M. tuberculosis ArgD that was then refined with Coot and PHENIX.

270 Structure validation was performed using Coot and PHENIX tools (33, 34). All figures were prepared

271 with PyMOL (The PyMOL Molecular Graphics System, Version 2.0 Schrodinger, LLC).

272

273 2.9. Isothermal titration calorimetry

274 Binding interaction between ArgB or ArgF and ligands was characterised at 25 °C, using a Microcal

275 ITC200 titration calorimeter (Microcal). An ArgB concentration between 75-150 µM was used for all

276 titrations. Ligands (0.75-2 mM) were injected in 1.5 µl aliquots with 150 s spacing between injections

277 for compound 1 and 110 s for all the others. For compound 2, L-canavanine and L-arginine two

278 titrations were concatenated. Titration data was recorded in 25 mM HEPES pH 7.4 with 200 mM NaCl.

279 Data were analysed by fitting a simple single-site model using Origin software (Microcal) (NMR711

280 and NMR446) or a six-site sequential binding model (ArgB and L-canavanine).

281 ArgF was dialysed in 50 mM HEPES pH 8.0, 200 mM NaCl before it was loaded into the calorimetry cell

282 at concentrations of 75-100 μM with the addition of 1 mM DTT. Ligand solutions at concentrations of

283 1 mM were dissolved in the same buffer and typically injected at between 0.5 μL and 2 μL at 150

284 second intervals with stirring at 750 rpm. Buffer-ligand titrations were carried out as reference runs

285 and subtracted from the protein-ligand titration to remove the heat of dilution. Data were analysed

286 by fitting a simple single-site model using Origin software (Microcal).

287

288 2.10. Enzymatic assays

289 ArgB activity was assessed by a colorimetric assay that followed the release of ADP by measuring the

290 oxidation of NADH at 340 nm, for 30 min, in the presence of pyruvate kinase and lactate

291 dehydrogenase in a PHERAstar plate-reader (BMG-Labtech). The enzymatic reactions (200 µl) were

292 performed at 25 °C and contained 50 mM Tris pH 7.5, 200 mM NaCl, 50 mM KCl, 10 mM MgCl2, 0.3

293 mM NADH, 2.5 mM phosphoenolpyruvate, 0.3 mM ATP, 1.25 mM N-acetyl-L-glutamate (NAG), 10%

294 DMSO (v/v), 4 units of pyruvate kinase/lactate dehydrogenase, 0.5 µM ArgB, and varying

295 concentration of inhibitors. Inhibitors were also individually screened against the coupled enzymes to

296 eliminate any compounds interfering with the other assay components. Competition assays were

297 performed in the same conditions using 0.3 mM of ATP or NAG and varying the other substrate

298 concentration.

299 To synthesise the ArgC substrate a reaction mixture containing 3 μM ArgB, 1 mM NAG and 1 mM ATP

300 in 50 mM Tris-HCl pH 7.5, 100 mM NaCl and 40 mM MgCl2 was made and the reaction was allowed to

301 proceed for 1-1.5 hours at room temperature. Thereafter, 100 μL of the ArgC reaction mixture

302 consisting of 50 mM Tris-HCl pH 7.5, 100 mM NaCl and 0.6 mM NADPH (concentration in 200 μL), was

303 added to each well together with 100 μL of the ArgB reaction mixture and followed for 30 min at 35

304 °C by measuring the oxidation of NADPH at 340 nm. Controls with only NADPH, 2.5% DMSO and no

305 ArgC, as well as only NADPH, 4 mM fragment and no ArgC were prepared. Baseline ArgC activity was

306 assayed with and without DMSO, the effect of 2.5% and 5% DMSO on enzymatic activity was also

307 tested. To test the inhibitory effect of fragment binders identified from the crystallographic screening,

308 6 fragment concentrations (125 μM, 250 μM, 500 μM, 1 mM, 2 mM and 4 mM,) were added to the

309 reaction mixtures from suitable DMSO stocks such that the final DMSO concentration was 2.5%. All

310 conditions were prepared in triplicates. 3.5 μM ArgC was added just before measurements were

311 started.

312 To assess the activity of ArgD an end point assay that follows the reverse reaction of the enzyme was

313 used in the presence and absence of fragments. Reactions (100 μL) were set up containing the

314 following: 100 mM Tris-HCl pH 8, 3 mM N-acetylornithine (NAO), 3.4 mM α-KG, 20 μM PLP and 4 μM

315 enzyme. Duplicate reaction mixtures were prepared for each time point and 8 time points (0, 5, 10,

316 15, 20, 30, 45, 60 minutes) were tested in total. Two controls were prepared: one with all reaction

317 components except the enzyme, and the other with all reaction components except NAO (NAO was

318 added after HCl treatment). The reactions tubes were allowed to equilibrate in a heating block set at

319 37 °C for two minutes, and the enzyme was added to initiate the reaction. After the stipulated

320 incubation times, contents of the reaction tubes were quickly transferred to 1.5 mL microcentrifuge

321 tubes containing 60 μL of 6 N HCl to stop the reactions. These tubes were then kept in a heating block

322 set at 95 °C for 30 minutes, after which they were cooled to 25 °C in a water bath. 200 μL of 3.6 M

323 sodium acetate was added to each tube (final concentration of 1.8 M), along with 40 μL of 30 mM 2-

324 aminobenzaldehyde (final concentration of 3 mM in a total volume of 400 μL). A yellow colouration

325 started developing as soon as the latter was added, the contents were vortexed and the tubes were

326 incubated at 25 °C for 15 minutes. 200 μL of each reaction mixture was transferred into wells of a 96-

327 well flat bottom UV transparent microplate, and absorbance at 440 nm was measured using the

328 PHERAstar plate reader. All experiments were performed at least in triplicate in a PHERAstar plate-

329 reader (BMG-Labtech) and the control without ArgD was used for blank subtraction. Data were

330 analysed with GraphPad Prism (Graphpad Software). All reagents were obtained from Sigma-Aldrich.

331

332 2.11. M. tuberculosis culture condition and minimum inhibitory concentration (MIC) determination

333 Mutant strain ∆argB and its complemented strain ∆argB-c were generated before as mentioned (10).

334 All the strains, wild type M. tuberculosis H37Rv and mutants ∆argB as well as ∆argB-c were grown at

335 37 °C to mid-log phase in Middlebrook 7H9 medium supplemented with 10% oleic acid-albumin-

336 dextrose-catalase (OADC), 0.5% glycerol, and 0.05% Tyloxapol supplemented with arginine (1mM)

337 washed and suspended in 7H9 media ± arginine (1mM). For MIC, the cultures were diluted to 1/500

338 in ± arginine supplemented media. Serial two-fold dilutions of each drug were prepared directly in a

339 sterile 96-well plate using 0.1 ml of media with the appropriate supplement in the presence or absence

340 of 1mM arginine. Same media with only vehicle (no drug) was used as a control. PBS (0.2 ml) was

341 added to all the perimeter wells. The diluted M. tuberculosis strains in ± arginine supplemented media

342 (0.1 ml) were added to each well, and the plate was incubated at 37°C for 7 days. Cell growth was

343 measured by optical density at 600 nm. An aqueous solution of resazurin (0.2 mg/ml; 0.03 ml) was

344 added to each well, and the plate was further incubated for up to two days at 37°C. The MIC was

345 determined as the lowest concentration at which the change of colour from blue (resazurin) to pink

346 (resorufin) did not occur.

347

348 3. Results

349 3.1. DSF fragment screening

350 In the first stage of the screening, differential scanning fluorimetry was used to screen an in-house

351 library of 960 rule-of-three compliant fragments against four enzymes of the arginine biosynthesis

352 pathway ArgB, ArgC, ArgD and ArgF. In the case of ArgB, ArgC and ArgF the screening was performed

353 against the apoenzymes while for ArgD it was done with the PLP-bound form. Several known ligands

354 (substrates, products, allosteric regulators and co-factors) were used as positive controls for each of

355 the enzymes. In all cases, a fragment was considered a hit when the shift in melting temperature was

356 greater than five times the standard deviation.

357

358 3.1.1. ArgB

359 In the conditions used in the assay, ArgB with 5% DMSO displayed a melting temperature of 48 °C. The

360 addition of 1 mM ATP, N-acetyl-glutamate and L-arginine showed positive melting shifts of 2.0, 11.8

361 and 8.6 °C respectively. Of the 960 compounds, a total of 63 (≈6.6%) showed a thermal shift greater

362 than five standard deviations (≥ 1.25 °C) at 5 mM and were considered hits. Out of those, 14 showed

363 a large stabilization of ArgB with a thermal shift greater than 5 °C (Table 1) and were selected for

364 further validation.

365

366 3.1.2. ArgC

367 The apoenzyme ArgC with 5% DMSO was found to have a melting temperature of 67.7 °C. While 1 mM

368 NADP+ gave a small average positive thermal shift of ~0.6 °C and 1 mM NADPH produced an average

369 negative thermal shift of -0.9 °C across all DSF runs performed. 81 out of the 960 fragments screened

370 (≈8.5% of the library) at a concentration of 5 mM gave a positive thermal shift greater than five

371 standard deviations (≥ 2.9 °C). An orthogonal biophysical technique, Surface Plasmon Resonance

372 (SPR), was also employed to corroborate these hits (Table 2).

373

374 3.1.3. ArgD

375 ArgD contains the prosthetic group PLP that does not leave the active site of the enzyme. Therefore,

376 all the PLP sites must remain occupied during screening experiments to represent the native state of

377 the target. In an attempt to saturate all of the ArgD PLP sites, PLP was added to ArgD during the

378 purification of the protein. Further confirmation was required to assess if most of the sites were now

379 saturated. To examine this, the melting profiles of unsaturated ArgD, saturated ArgD and the two

380 forms in the presence 1 mM PLP were assessed. Unsaturated ArgD exhibited a melting profile with 2

381 peaks, a large and broad peak with melting temperature of 64.5 °C, and a small peak with a melting

382 temperature of 77 °C (Figure S1). The addition of 1 mM PLP changed the melting profile to a single

383 peak with a melting temperature to 77.5 °C (Figure S1). ArgD that had PLP added during purification

384 showed a melting temperature of 77.5 °C (Figure S1). The addition of 1 mM PLP had now a very minor

385 stabilizing effect shifting the melting temperature to 79 °C (Figure S1) and confirming that most PLP

386 sites were saturated. However, the protein in this state was insensitive to fragment binding with

387 maximum thermal shifts of 0.5 °C being observed. We therefore tested the potential of using the

388 unsaturated protein for the screening. As mentioned above, PLP-unsaturated ArgD exhibited a melting

389 profile with 2 peaks, a large and broad peak with melting temperature of 64.5 °C (Figure S1), likely a

390 mixture of two different populations in which none or one of the two protomers contain PLP, and a

391 small peak with a melting temperature of 77 °C (Figure S1) which most likely corresponded to a PLP-

392 saturated population. The fact that the addition of 1 mM PLP shifts the melting profile to a single peak

393 with a melting temperature of 77.5 °C corroborated this hypothesis.

394 Three different types of response in the melting profile to the presence of fragments were observed

395 while screening the PLP-unsaturated ArgD. Most fragments showed either no effect on the melting

396 profile or a decrease in the melting temperature of the large peak or of both peaks and were discarded

397 (Figure S1). A second set caused a change in the melting profile with a large increase in the intensity

398 of the highest temperature peak (Figure S1) suggesting that the fragment was preferentially binding

399 to the PLP binding site. The third set, had fragments that led to an increase in melting temperature

400 inferior to 4 °C but maintained the melting profile of the PLP-unsaturated ArgD control (Figure S1).

401 Fragments of this set could either be binding to the PLP site but not stabilizing the protein sufficiently

402 to show a clear change in the melting profile, or could be binding elsewhere on the protein both in

403 the presence or absence of PLP. This set, represented by 47 fragments (≈4.9% of the library) giving

404 melting shifts of at least five standard deviations (≥ 2 °C) were therefore selected for SPR validation

405 (Table 3).

406

407 3.1.4. ArgF

408 ArgF displayed a melting temperature of 65.5 °C with 1 mM L-ornithine and L-citrulline showing a

409 positive melting shift of 1.5 and 1 °C respectively. Of the 960 fragments, a total of 105 (≈10.9%) showed

410 a thermal shift greater than five times standard deviations (≥ 1.0 °C) at 5 mM and were considered

411 hits. Out of these 16 displayed a melting shift greater than 3 °C and were selected for further validation

412 by X-ray. Due to the large number of hits for this protein, greater than 10% of the whole library a

413 clustering analysis of the fragment hits was performed. Centroids for each identified cluster and the

414 representative displaying the highest melting shift were also selected for X-ray validation (Table 4).

415

416 3.2. Validation of DSF hits using a secondary screening technique

417 Three different strategies were employed to validate the hits obtained with DSF. For ArgF the

418 fragment hits were taken directly for X-ray crystallographic validation, for ArgB ligand-based NMR was

419 employed and for ArgC and ArgD SPR was performed to validate the DSF hits.

420

421 3.2.1. ArgB

422 To validate the ArgB hits obtained by DSF two ligand based NMR methods, Carr-Purcell-Meiboom-Gill

423 (CPMG) and STD (saturated transfer difference), were used (36, 37) and fragments that were validated

424 by at least one method were considered as confirmed hits. CPMG experiments validated 15 out of the

425 16 fragments while CPMG validated 11 out of 16. Only one fragment was not validated by both

426 methods and thus was not taken forward.

427

428 3.2.2. ArgC and ArgD

429 ArgC and PLP saturated ArgD were immobilised on an activated carboxymethylated dextran surface

430 via amine coupling and final immobilisation response achieved was ≈16000 RU for ArgC and ≈7500 RU

431 for ArgD (1000 RU roughly corresponding to 10 mg/mL protein immobilised on the surface). For ArgD,

432 a 1 mM PLP injection after immobilization increased the absolute baseline immobilisation response

433 by ≈1100 RU. This was done to compensate for loss of PLP in the low pH coupling buffer during

434 immobilisation. The baseline throughout the screening experiment remained at the post-PLP injection

435 level (≈8570 RU), indicating that PLP was not lost during the experiment.

436 The immobilised ArgC protein was tested first using dilution series of NADPH and NADP+ (19 μM to 2.5

437 mM), and a clear dose response suggested that predominantly, the enzyme had not been immobilised

438 in an orientation that occluded the active site. Similarly, ArgD was tested using a dilution series of L-

439 glutamate, N-acetylornithine and L-ornithine (19 μM to 2.5 mM).

440 The fragment hits obtained previously were injected at a concentration of 1 mM. The sensorgrams for

441 all the analytes were inspected visually to exclude fragments with either no discernible response or a

442 “sticky” profile from further analysis. The binding level was calculated using the T200 software and

443 adjusted for molecular weights of the analytes. For ArgC, 22 fragments with a binding level response

444 ≥20 RU were shortlisted, whereas in the case of ArgD, 20 fragments with a response ≥10 RU were

445 shortlisted for crystallographic validation. Fragments were thereafter described either as SPR

446 ‘positive’ or ‘negative’.

447

448

449

450

451 3.3. Crystallographic, biophysical and biochemical validation

452 The hits obtained for ArgB, ArgC, ArgD and ArgF were then soaked in crystals of the respective protein

453 and X-ray diffraction data was collected. Data collection and refinement statistics for all structures are

454 available in Table S3.

455

456 3.3.1. ArgC

457 Crystal structures were obtained for the ArgC apoenzyme (PDB: 7NNI) and the NADP+-bound

458 holoenzyme (PDB: 7NNQ) (Figure 2A). Binding of NAPD causes significant structural changes in two

459 loops of the protein that move from a closed to open conformation (Figure 2A). Additionally,

460 structures were solved with 4 fragment binders, occupying either of the two distinct pockets: the

461 substrate-binding (Figure 2C and D) and the NADP(H)-binding pockets (Figure 2E and F). Fragments

462 NMR322 (5-Methoxy-3-indoleacetic acid) and NMR 571 (Xanthene-9-carboxylic acid) were observed

463 in the substrate-binding pocket (PDB: 7NOT and 7NNR respectively) (Figure 2C and D). Both NMR322

464 and NMR571 engaged side chains of residues His217 and Tyr211 through hydrogen bonds. Both

465 residues are predicted to stabilise the acyl-enzyme intermediate during catalysis. NMR322 also made

466 an H-bond interaction with ser186 and gly187. As compared to NMR571, NMR322 binds deeper in the

467 pocket (Figure 2C and D).

468 Fragment NMR401 (6-phenoxy-3-pyridinamine) and NMR863 (5-methoxy-1,3-benzoxazole-2-

469 carboxylic acid) were observed in the ribosylnicotinamide and pyrophosphate regions of the NADP(H)-

470 binding pockets (PDB: 7NPJ and 7NPH respectively) and the majority of the interactions between the

471 protein and these two fragments are hydrophobic or π-interactions. Both fragments form only a single

472 hydrogen bond interaction with the thr325 side chain in case of NMR401 (Figure 2E) and with the

473 arg193 backbone amine in the case of NMR863.

474 Enzymatic assays revealed that 2 mM NMR322 inhibited ArgC activity by 45% whereas 2 mM NMR571

475 caused a 37% inhibition. 2 mM NMR401 inhibited ArgC activity by 10% whereas 2 mM NMR863 caused

476 a 12% inhibition. Although the thermal shifts obtained for NADP(H)-binding pocket fragments

477 NMR401 and NMR863 were higher than those for substrate-binding pocket fragments NMR322 and

478 NMR571, the SPR binding response for the latter was better and positively correlated with percentage

479 inhibition of enzymatic activity (Table 2).

480

481 3.3.2. ArgD

482 The first crystal structure of the ArgD holoenzyme from M. tuberculosis obtained (Figure 3A) showed

483 that the protomer has three domains: the smaller N-terminal segment (residues 7 to 74), the relatively

484 larger C-terminal domain (residues 286 to 396), and the central PLP-binding domain (residues 85 to

485 273), which is also the largest and has a Rossmann-like overall fold (Figure 3B). The prosthetic group

486 PLP is covalently linked to Lys253 via an aldimine linkage. ArgD is a dimeric enzyme like other members

487 of the class III δ-aminotransferase family (38); the active sites are interfacial, and residues of both

488 protomers contribute to the active site architecture (Figure 3A) (PDB: 7NN1).

489 Fragment NMR608 (3-Hydroxy-2-naphthoic acid) was observed occupying all four binding sites (4

490 chains in the ASU). It binds closer to the distal phosphate group of PLP than the proximal ring system

491 and the internal aldimine bond, where catalysis actually occurs. However, it makes, through its acid

492 group, hydrogen bond interactions with Arg142, which could be the key residue involved in substrate

493 binding based on homology with AstC, and with an highly coordinated water through the hydroxyl

494 group (Figure 3C). While the pocket itself has depth, the fragment is more solvent-exposed than the

495 PLP molecule. While NMR608 was the only hit validated by all techniques the fragment library

496 contained a compound with similar structure (NMR868) that had only a melting shift of 0.9 C and

497 therefore wasn’t selected initially for further testing. However, soaking of this fragment revealed that

498 it binds in almost the same position as NMR608 (Figure 3D). Nevertheless, The orientation of the acid

499 group is slightly different in NMR868 and it no longer interacts directly with arg142. In fact this

500 fragment only has hydrogen bond interactions with solvent molecules and it is only observed in one

501 molecule in the ASU out of four, most likely reflecting a lower affinity (PDB: 7NNC).

502 Aminotransferases are often assayed for the reverse reactions they catalyse because in most cases

503 substrates for the forward reaction are not commercially sold. The ArgD holoenzyme can also use N-

504 acetylornithine (NAO) and α-ketoglutarate (α-KG) to generate N-acetyl-γ-glutamyl-semialdehyde and

505 glutamate. The semialdehyde product spontaneously cyclises into Δ1-pyrroline-5-carboxylic acid and

506 can react with the reagent 2-aminobenzaldehyde to yield a dihydroquinazolinium compound (bright

507 yellow colouration) that absorbs at 440 nm (39, 40). This assay was employed to assess the effect of

508 the fragment hits on on ArgD. NMR608 exhibited very weak activity with only 31% and 16% inhibition

509 observed at 4 and 2mM respectively. This is consistent with the crystal structures where the fragment

510 is highly exposed to the solvent.

511

512 3.3.3. ArgF

513 Crystal structures of apo ArgF and of ArgF in complex with the natural ligand carbamoyl phosphate

514 were initially obtained (PDB: 7NNF and 7NNV respectively). Carbamoyl phosphate interacts with ArgF

515 through hydrogen bonds with the side chains of ser50, thr51, arg52, arg101, his128 and gln131 but

516 also with the backbone amines of thr51, arg52 and thr53 and the carbonyl of cys264 (Figure 4A). The

517 residues ser50, thr51, arg52 are at the N-terminus of α-helix 2 and the phosphate group of carbamoyl

518 phosphate sit at the positive pole of the helix. Binding of carbamoyl phosphate to ArgF slightly

519 displaces α-helix 2 when compared to the apo structure (Figure S2A). Soaking ArgF with fragment hits

520 yielded eight crystals structures. All fragments occupied a site at the interface between two protomers

521 of the ArgF trimer but not all sites are equally occupied by all fragments (Figure 4B-E and S2B-F). This

522 site contacts directly with the α-helix 2 and sits between this helix and α-helix 3 of the opposing

523 protomer. The site is formed by residues thr51, arg52, phe55 of α-helix 2, leu265, ala289, arg292 of

524 protomer 1 and ile45, ser75, thr76, leu78, glu82, thr87, leu91 and tyr94 of protomer 2. These residues

525 form a cavity that opens to the carbamoyl phosphate binding site. Binding of fragments at this site,

526 slightly shifts the position of α-helix 2 by 1.9 Å, when compared to the carbamoyl phosphate structure,

527 and also affects the conformation of arg52 which is involved in carbamoyl phosphate binding.

528 Fragments were found to bind in two distinct sub sites across the main binding site and could be

529 divided in three different groups based on their mode of binding. NMR007, 078, 464 and 502 occupied

530 the top area of the site (subsite 1) and contacted with a loop of protomer 2 that is composed by

531 residues asp72 to leu84 that covers the site (Figure 4C and S2B-D). NMR801, the single representative

532 of this group sits at the bottom of the site (subsite 2) between α-helix 2 of protomer 1 and α-helix 3

533 of the opposing protomer (Figure 4D). NMR288, 812 and 817 have two molecules binding at this site,

534 with one molecule occupying each subsite (Figure 4E and S2E-F). All fragments keep α-helix 2 in a

535 position close to the apo structure or move further away from the position this helix occupies when

536 carbamoyl phosphate is bound, albeit very slightly. A Kd value could only be determined for NMR007

537 and NMR812 and both fragments showed affinities worse than 100 µM (Table 4 and Figure S3).

538

539 3.3.4. ArgB

540 We obtained crystal structures for ArgB in the apo form and with the natural ligands N-acetyl-

541 glutamate (NAG) and L-arginine (PDB: 7NLF, 7NLN, and 7NLO respectively) (Figure 5A). The thirteen

542 crystal structures of ArgB with fragments in the absence of natural ligands show that all the

543 compounds were unexpectedly bound to a hydrophobic cavity at the interface between two

544 protomers with three of these sites present in the ArgB hexamer (Fig. 5B-D and Figure S4). L-

545 canavanine, the guanidinooxy structural analogue of L-arginine, bound to the same site of ArgB as L-

546 arginine and induced conformational changes similar to those induced by L-arginine (Figure S5). The

547 interfacial site is composed by ala124, val125, gly126, ile127, asp131, ala132, leu134, ala164, met165,

548 leu168 and arg173 and is mostly hydrophobic in nature (Figure S6). Due to the nature of this new site,

549 the interactions between the compounds and the protein are essentially hydrophobic, with residues

550 leu168 and val125 interacting with NMR711 (PDB: 7NNB) while for NMR446 (PDB: 7NLX) ile127 is also

551 involved in the hydrophobic interactions (Fig. 5C and D). Carbon-π interactions are also formed with

552 leu168 for both compounds (Fig. 5C and D). Finally, weak hydrogen bonds are also present between

553 asp167, leu168 and val125 for NMR711 while NMR446 interacts with ile127 and leu168 via

554 hydrophobic contacts (Fig. 5C and D). Furthermore, this site is symmetrical and sits at a 2-fold

555 crystallographic symmetry axis with each compound clearly presenting two binding conformations

556 (Figure 5E and S4). The two compounds (NMR446 and NMR711) also share structural and binding

557 features with a trifluoromethyl group occupying the same position at the binding site.

558 Enzymatic assays show that, of the 15 compounds, NMR711 [2,8-bis(trifluoromethyl)-1H-quinolin-4-

559 one] and NMR446 [8-(trifluoromethyl)-1H-quinolin-4-one] were the best ArgB inhibitors, with an IC50

560 of 366 and 707 µM, respectively (Fig. 6A and Table 1). The natural allosteric regulator L-arginine and

561 its analogue L-canavanine (Fig. 6A) have IC50 values of 186 µM and 1.46 mM respectively. Additionally,

562 ITC experiments showed that compounds NMR711 and NMR446 bind to ArgB with a KD of 7.7 and 23

563 µM, respectively (Table 1 and Figure S7), whereas L-arginine and L-canavanine showed complex

564 binding curves that can only be fitted to a sequential binding model showing a cooperative interaction

565 with the different protomers of the hexamer (Table 1 and Figure S7). NMR competition assays

566 revealed that compounds NMR711 and NMR446 are not competitive with any of the natural ligands

567 (ATP, NAG, and L-arginine), with enzymatic assays also demonstrating the non-competitive nature of

568 the inhibition of both fragments. This confirms that the results from X-ray crystal structures are not

569 an artefact and that the fragment-binding site is indeed a new allosteric site (Fig. 6B and S8). The

570 observation that there are no conformational changes in the crystal structures of ArgB with either

571 fragment at the obtained resolution (2-2.5 Å) may be due to constraints arising from crystal packing.

572 Bioinformatics analysis showed that this site is conserved in mycobacterial species and also in closely

573 related actinobacteria, such as Nocardia (Figure S6B). Nevertheless, it is clear that binding of these

574 compounds, similar to L-arginine binding, causes changes in the energy landscape of the protein that

575 result is allosteric inhibition of the catalytic reaction.

576

577 3.4. Effect of ArgB inhibitors in M. tuberculosis growth

578 Considering the four enzymes screened of the arginine biosynthesis pathway, ArgB hits exhibited

579 higher potency by far, with NMR711 and NMR446 being selected to assess their effect on M.

580 tuberculosis together with L-canavanine.

581 The ability of these compounds to inhibit M. tuberculosis growth was examined by measuring their

582 minimum inhibitory concentrations (MICs) in the absence or presence of arginine (1 mM). All the

583 compounds inhibited the growth of M. tuberculosis H37Rv and ∆argB-c in media without arginine

584 compared to no drug control (Figure 7). MICs for NMR711, NMR446 and L-canavanine were 25-50,

585 >200 and 50 µg/ml against H37Rv and ∆argB-c (Figure 7). However, when arginine (1 mM) was present

586 in the media, compound NMR446 and L-canavanine had no inhibitory activity against H37Rv, ∆argB-c

587 and, ∆argB. This indicates that these compounds are indeed specifically inhibiting M. tuberculosis

588 arginine biosynthesis (Figure 7). In contrast, the more promiscuous NMR711 was inhibitory for all the

589 above strains in the presence or absence of arginine, suggesting that NMR711 may target additional

590 proteins (Figure7).

591 Using the FBDD approach, we have discovered inhibitors that bind to a new allosteric site in ArgB,

592 which has very different properties than that of the active site and L-arginine binding sites, thus

593 opening new possibilities for drug discovery by targeting ArgB. For fragment-sized molecules, both

594 compounds reported in this work bind tightly and allosterically to ArgB and have growth inhibitory

595 activity against M. tuberculosis, suggesting that they have the potential to provide a framework for

596 developing larger and higher affinity molecules against the ArgB protein.

597

598 4. Discussion

599 The arginine biosynthesis pathway has been established as a good target for anti-TB drug discovery

600 (10, 14). Arginine deprivation in M. tuberculosis induced by knocking out argB and argF results in both

601 in vitro and in vivo sterilisation of M. tuberculosis, without the emergence of suppressor mutants (10).

602 However, from a pathway with eight enzymes, only ArgJ has been explored in a drug discovery

603 campaign and all other enzymes of the pathway, prior to this work, were yet to be assessed in their

604 potential as suitable targets for drug discovery.

605 Fragments are potent chemical tools that can efficaciously explore the surface of proteins for new

606 binding sites and their chemical space, even with small libraries of a few hundreds of compounds and

607 can therefore be employed to assess the ligandability of protein targets (18, 41). Therefore, this

608 approach was employed to assess the ligandability of ArgB, ArgC, ArgD and ArgF, to identify potential

609 starting points for fragment development.

610 We have screened a fragment library of 960 small compounds (MW 150-300 Da) initially using DSF

611 and employed ligand-based NMR, SPR, ITC, biochemical assays and X-ray crystallography to validate

612 the hits. Due to the nature of FBDD a hit is only considered validated when an X-ray crystal structure

613 is obtained. For all the proteins in this work, hits were found and eventually validated by X-ray

614 crystallography. ArgB had the highest number of X-ray validated hits with a total of fourteen, followed

615 by ArgF with eight, ArgC with four and ArgD with two. Interestingly, in the case of ArgB and ArgF, all

616 the fragment hits were binding to an interfacial site, which in the case of ArgB was confirmed to have

617 functional implications. In the case of ArgF, its close proximity to the active site shows potential to

618 develop compounds that can anchor at the interfacial site and then extend towards the active site,

619 thus inhibiting the enzyme. Similarly for ArgJ, the only enzyme of the pathway with known inhibitors

620 prior to this work, the inhibitors were also found to bind to an interfacial allosteric site (14). Our results

621 further show that in the case of ArgC there are two possible strategies to develop inhibitors, with one

622 targeting the cofactor binding site and the other the substrate binding site. It is not clear at this point

623 which strategy has the highest potential to result in potent inhibitors. Another consideration to take

624 into account is the level of homology of these enzymes with the human orthologue. While ArgB and

625 ArgC do not have a human orthologue, ArgD and ArgF do and the M. tuberculosis enzymes have

626 identities of 36 and 41% with the human orthologues, respectively. Nevertheless, while the ArgF active

627 site is conserved, the interfacial site of ArgF contains several differences that raise the prospect of

628 developing specific inhibitors for the M. tuberculosis enzyme. For ArgD selectivity might be more

629 difficult to achieve since many of the active site residues are conserved in comparison with the human

630 cytoplasmic and mitochondrial enzymes.

631 Due to the potency of the best fragments against ArgB, we tested them for their ability to inhibit M.

632 tuberculosis growth together with L-canavanine. Remarkably, NMR446 and L-canavanine not only

633 inhibited M. tuberculosis growth, but were also found to act on-target despite the potential

634 promiscuity of such small compounds, with both becoming inactive after the addition of L-arginine to

635 the media.

636 Despite these promising results, the interfacial site of ArgB might be the hardest of all sites found in

637 this study to develop small molecule inhibitors. The intrinsic highly hydrophobic nature of the site

638 together with very few opportunities to engage in hydrogen bonds and other polar contacts creates

639 difficulties in rationalizing what modifications could improve the potency of the compounds.

640 Furthermore, the fact that we cannot observe conformational changes in any of the structures with

641 fragments bound to ArgB may be due to constraints arising from crystal packing and thus these

642 structures may not completely represent what happens in solution. It is however also possible that

643 binding to this site does not cause visible conformational changes but still alters the energy landscape

644 of the intramolecular pathways involved in the catalytic cycle. We cannot currently determine which

645 of these two hypotheses is correct.

646 In conclusion, using a fragment-based approach, we have discovered inhibitors that bind to novel sites

647 in ArgB and ArgF and to the active sites of ArgC and ArgD, which in case of ArgB show on target activity

648 against M. tuberculosis. The data presented here clearly shows that there is scope to target at least

649 ArgC and ArgF with dedicated drug discovery programs and we propose these two as the best

650 candidates for future drug discovery work.

651

652 References

653 1. World Health O. 2019. Global tuberculosis report 2019. World Health Organization, Geneva. 654 2. Vilcheze C, Jacobs WR, Jr. 2019. The Isoniazid Paradigm of Killing, Resistance, and Persistence 655 in Mycobacterium tuberculosis. J Mol Biol 431:3450-3461. 656 3. Berney M, Berney-Meyer L, Wong KW, Chen B, Chen M, Kim J, Wang J, Harris D, Parkhill J, 657 Chan J, Wang F, Jacobs WR, Jr. 2015. Essential roles of methionine and S-adenosylmethionine 658 in the autarkic lifestyle of Mycobacterium tuberculosis. Proc Natl Acad Sci U S A 112:10008- 659 13. 660 4. Gouzy A, Poquet Y, Neyrolles O. 2014. Nitrogen metabolism in Mycobacterium tuberculosis 661 physiology and virulence. Nat Rev Microbiol 12:729-37. 662 5. Zhang YJ, Rubin EJ. 2013. Feast or famine: the host-pathogen battle over amino acids. Cell 663 Microbiol 15:1079-87. 664 6. Lee S, Jeon BY, Bardarov S, Chen M, Morris SL, Jacobs WR, Jr. 2006. Protection elicited by two 665 glutamine auxotrophs of Mycobacterium tuberculosis and in vivo growth phenotypes of the 666 four unique glutamine synthetase mutants in a murine model. Infect Immun 74:6491-5. 667 7. Sambandamurthy VK, Wang X, Chen B, Russell RG, Derrick S, Collins FM, Morris SL, Jacobs WR, 668 Jr. 2002. A pantothenate auxotroph of Mycobacterium tuberculosis is highly attenuated and 669 protects mice against tuberculosis. Nat Med 8:1171-4. 670 8. Smith DA, Parish T, Stoker NG, Bancroft GJ. 2001. Characterization of auxotrophic mutants of 671 Mycobacterium tuberculosis and their potential as vaccine candidates. Infect Immun 69:1142- 672 50. 673 9. Hondalus MK, Bardarov S, Russell R, Chan J, Jacobs WR, Jr., Bloom BR. 2000. Attenuation of 674 and protection induced by a leucine auxotroph of Mycobacterium tuberculosis. Infect Immun 675 68:2888-98. 676 10. Tiwari S, van Tonder AJ, Vilcheze C, Mendes V, Thomas SE, Malek A, Chen B, Chen M, Kim J, 677 Blundell TL, Parkhill J, Weinrick B, Berney M, Jacobs WR, Jr. 2018. Arginine-deprivation- 678 induced oxidative damage sterilizes Mycobacterium tuberculosis. Proc Natl Acad Sci U S A 679 doi:10.1073/pnas.1808874115. 680 11. Peteroy-Kelly MA, Venketaraman V, Talaue M, Seth A, Connell ND. 2003. Modulation of J774.1 681 macrophage L-arginine metabolism by intracellular Mycobacterium bovis BCG. Infect Immun 682 71:1011-5. 683 12. Seth A, Connell ND. 2000. Amino acid transport and metabolism in mycobacteria: cloning, 684 interruption, and characterization of an L-Arginine/gamma-aminobutyric acid permease in 685 Mycobacterium bovis BCG. J Bacteriol 182:919-27. 686 13. Gobert AP, Daulouede S, Lepoivre M, Boucher JL, Bouteille B, Buguet A, Cespuglio R, Veyret B, 687 Vincendeau P. 2000. L-Arginine availability modulates local nitric oxide production and 688 parasite killing in experimental trypanosomiasis. Infect Immun 68:4653-7.

689 14. Mishra A, Mamidi AS, Rajmani RS, Ray A, Roy R, Surolia A. 2018. An allosteric inhibitor of 690 Mycobacterium tuberculosis ArgJ: Implications to a novel combinatorial therapy. EMBO Mol 691 Med 10. 692 15. DeJesus MA, Gerrick ER, Xu W, Park SW, Long JE, Boutte CC, Rubin EJ, Schnappinger D, Ehrt S, 693 Fortune SM, Sassetti CM, Ioerger TR. 2017. Comprehensive Essentiality Analysis of the 694 Mycobacterium tuberculosis Genome via Saturating Transposon Mutagenesis. MBio 8. 695 16. Erlanson DA, Fesik SW, Hubbard RE, Jahnke W, Jhoti H. 2016. Twenty years on: the impact of 696 fragments on drug discovery. Nat Rev Drug Discov 15:605-19. 697 17. Thomas SE, Collins P, James RH, Mendes V, Charoensutthivarakul S, Radoux C, Abell C, Coyne 698 AG, Floto RA, von Delft F, Blundell TL. 2019. Structure-guided fragment-based drug discovery 699 at the synchrotron: screening binding sites and correlations with hotspot mapping. Philos 700 Trans A Math Phys Eng Sci 377:20180422. 701 18. Mendes V, Blundell TL. 2017. Targeting tuberculosis using structure-guided fragment-based 702 drug design. Drug Discov Today 22:546-554. 703 19. Thomas SE, Whitehouse AJ, Brown K, Burbaud S, Belardinelli JM, Sangen J, Lahiri R, Libardo 704 MDJ, Gupta P, Malhotra S, Boshoff HIM, Jackson M, Abell C, Coyne AG, Blundell TL, Floto RA, 705 Mendes V. 2020. Fragment-based discovery of a new class of inhibitors targeting 706 mycobacterial tRNA modification. Nucleic Acids Res 48:8099-8112. 707 20. Sabbah M, Mendes V, Vistal RG, Dias DMG, Zahorszka M, Mikusova K, Kordulakova J, Coyne 708 AG, Blundell TL, Abell C. 2020. Fragment-Based Design of Mycobacterium tuberculosis InhA 709 Inhibitors. J Med Chem 63:4749-4761. 710 21. Whitehouse AJ, Thomas SE, Brown KP, Fanourakis A, Chan DS, Libardo MDJ, Mendes V, 711 Boshoff HIM, Floto RA, Abell C, Blundell TL, Coyne AG. 2019. Development of Inhibitors against 712 Mycobacterium abscessus tRNA (m(1)G37) Methyltransferase (TrmD) Using Fragment-Based 713 Approaches. J Med Chem 62:7210-7232. 714 22. Villemagne B, Machelart A, Tran NC, Flipo M, Moune M, Leroux F, Piveteau C, Wohlkonig A, 715 Wintjens R, Li X, Gref R, Brodin P, Deprez B, Baulard AR, Willand N. 2020. Fragment-Based 716 Optimized EthR Inhibitors with in Vivo Ethionamide Boosting Activity. ACS Infect Dis 6:366- 717 378. 718 23. Prati F, Zuccotto F, Fletcher D, Convery MA, Fernandez-Menendez R, Bates R, Encinas L, Zeng 719 J, Chung CW, De Dios Anton P, Mendoza-Losana A, Mackenzie C, Green SR, Huggett M, Barros 720 D, Wyatt PG, Ray PC. 2018. Screening of a Novel Fragment Library with Functional Complexity 721 against Mycobacterium tuberculosis InhA. ChemMedChem 13:672-677. 722 24. Villemagne B, Flipo M, Blondiaux N, Crauste C, Malaquin S, Leroux F, Piveteau C, Villeret V, 723 Brodin P, Villoutreix BO, Sperandio O, Soror SH, Wohlkonig A, Wintjens R, Deprez B, Baulard 724 AR, Willand N. 2014. Ligand efficiency driven design of new inhibitors of Mycobacterium 725 tuberculosis transcriptional repressor EthR using fragment growing, merging, and linking 726 approaches. J Med Chem 57:4876-88. 727 25. Ribeiro JA, Hammer A, Libreros-Zuniga GA, Chavez-Pacheco SM, Tyrakis P, de Oliveira GS, 728 Kirkman T, El Bakali J, Rocco SA, Sforca ML, Parise-Filho R, Coyne AG, Blundell TL, Abell C, Dias 729 MVB. 2020. Using a Fragment-Based Approach to Identify Alternative Chemical Scaffolds 730 Targeting Dihydrofolate Reductase from Mycobacterium tuberculosis. ACS Infect Dis 6:2192- 731 2201. 732 26. Faion L, Djaout K, Frita R, Pintiala C, Cantrelle FX, Moune M, Vandeputte A, Bourbiaux K, 733 Piveteau C, Herledan A, Biela A, Leroux F, Kremer L, Blaise M, Tanina A, Wintjens R, Hanoulle 734 X, Deprez B, Willand N, Baulard AR, Flipo M. 2020. Discovery of the first Mycobacterium 735 tuberculosis MabA (FabG1) inhibitors through a fragment-based screening. Eur J Med Chem 736 200:112440. 737 27. Peranen J, Rikkonen M, Hyvonen M, Kaariainen L. 1996. T7 vectors with modified T7lac 738 promoter for expression of proteins in Escherichia coli. Anal Biochem 236:371-3.

739 28. Mendes V, Acebron-Garcia-de-Eulate M, Verma N, Blaszczyk M, Dias MVB, Blundell TL. 2019. 740 Mycobacterial OtsA Structures Unveil Substrate Preference Mechanism and Allosteric 741 Regulation by 2-Oxoglutarate and 2-Phosphoglycerate. mBio 10. 742 29. Cherney LT, Cherney MM, Garen CR, Niu C, Moradian F, James MN. 2007. Crystal structure of 743 N-acetyl-gamma-glutamyl-phosphate reductase from Mycobacterium tuberculosis in complex 744 with NADP(+). J Mol Biol 367:1357-69. 745 30. Vonrhein C, Flensburg C, Keller P, Sharff A, Smart O, Paciorek W, Womack T, Bricogne G. 2011. 746 Data processing and analysis with the autoPROC toolbox. Acta Crystallogr D Biol Crystallogr 747 67:293-302. 748 31. Winter G. 2010. xia2: an expert system for macromolecular crystallography data reduction. 749 Journal of Applied Crystallography 43:186-190. 750 32. McCoy AJ, Grosse-Kunstleve RW, Adams PD, Winn MD, Storoni LC, Read RJ. 2007. Phaser 751 crystallographic software. J Appl Crystallogr 40:658-674. 752 33. Adams PD, Afonine PV, Bunkoczi G, Chen VB, Davis IW, Echols N, Headd JJ, Hung LW, Kapral 753 GJ, Grosse-Kunstleve RW, McCoy AJ, Moriarty NW, Oeffner R, Read RJ, Richardson DC, 754 Richardson JS, Terwilliger TC, Zwart PH. 2010. PHENIX: a comprehensive Python-based system 755 for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr 66:213-21. 756 34. Emsley P, Lohkamp B, Scott WG, Cowtan K. 2010. Features and development of Coot. Acta 757 Crystallogr D Biol Crystallogr 66:486-501. 758 35. Afonine PV, Grosse-Kunstleve RW, Echols N, Headd JJ, Moriarty NW, Mustyakimov M, 759 Terwilliger TC, Urzhumtsev A, Zwart PH, Adams PD. 2012. Towards automated crystallographic 760 structure refinement with phenix.refine. Acta Crystallogr D Biol Crystallogr 68:352-67. 761 36. Mayer M, Meyer B. 1999. Characterization of Ligand Binding by Saturation Transfer Difference 762 NMR Spectroscopy. Angew Chem Int Ed Engl 38:1784-1788. 763 37. Loria JP, Rance M, Palmer AG. 1999. A Relaxation-Compensated Carr−Purcell−Meiboom−Gill 764 Sequence for Characterizing Chemical Exchange by NMR Spectroscopy. Journal of the 765 American Chemical Society 121:2331-2332. 766 38. Newman J, Seabrook S, Surjadi R, Williams CC, Lucent D, Wilding M, Scott C, Peat TS. 2013. 767 Determination of the structure of the catabolic N-succinylornithine transaminase (AstC) from 768 Escherichia coli. PLoS One 8:e58298. 769 39. Vogel HJ, Davis BD. 1952. Glutamic γ-Semialdehyde and Δ1-Pyrroline-5-carboxylic Acid, 770 Intermediates in the Biosynthesis of Proline1,2. Journal of the American Chemical Society 771 74:109-112. 772 40. Albrecht AM, Vogel HJ. 1964. Acetylornithine Delta-Transaminase. Partial Purification and 773 Repression Behavior. J Biol Chem 239:1872-6. 774 41. Canning P, Birchall K, Kettleborough CA, Merritt A, Coombs PJ. 2020. Fragment-based target 775 screening as an empirical approach to prioritising targets: a case study on antibacterials. Drug 776 Discov Today doi:10.1016/j.drudis.2020.09.003. 777 42. Wang Q, Xia J, Guallar V, Krilov G, Kantrowitz ER. 2008. Mechanism of thermal decomposition 778 of carbamoyl phosphate and its stabilization by aspartate and ornithine transcarbamoylases. 779 Proc Natl Acad Sci U S A 105:16918-23. 780

781 Acknowledgements

782 This work was funded by Bill and Melinda Gates Foundation HIT-TB (OPP1024021) and SHORTEN-TB

783 (OPP1158806). PA was funded by a Gates Cambridge Scholarship. TLB is funded by the Wellcome Trust

784 (Wellcome Trust Investigator Award 200814_Z_16_Z: RG83114). The authors would like to thank the

785 Diamond Light Source for beam-time (proposals mx14043 and mx18548).

786

787 Competing Interest declaration

788 The authors declare no competing interests.

789

790

791

792

793

794

795

796

797

798

799

800

801

802

803

804

805

806

807

808

809

810 Tables

811

812 Table 1: ArgB validated fragment hits.

DSF ΔTm ITC IC50 (µM) Compound Fragment structure CPMG STD (°C) Kd (µM) ‡ # OH NMR026 +7.6 binds binds ND ND N O H N no NMR043 +6.6 binds ND ND binding N O I NMR078 O +6.1 binds binds ND ND

HO O NMR082 N N +7.0 binds binds ND ND OH

HN O NMR314 +9.4 binds binds ND ND

H N NMR323 +5.4 binds binds ND ND

O CN F3C N NMR446 +8.0 binds binds 24 ± 1.5 707 ± 7

OH H N no NMR462 +6.0 binds ND ND Cl binding N H F3C N NMR469 +10.5 binds binds ND ND O OH O

NMR582 +6.3 binds binds ND ND

OH O O

O O no NMR612 +5.3 binds ND ND binding

H N no NMR617 +6.4 binds ND ND binding CN CF3 N CF3 NMR711 +7.9 binds binds 7.7 ± 0.8 366 ± 4

OH L-arginine +11.8 - - * 146 ± 1 NAG +8.6 - - 56 ± 5 - ATP +2.0 - - ND - L-canavanine +7.2 - - * 1460 ± 2 813 ND - not determined.

814 ‡ - Attempts were made to determine the Kd for these ligands but without success.

815 * - L-arginine and L-canavanine ITC data could only be fitted with a sequential binding and the best fit

816 was a six site model with Kd of 4.5 ± 1.0, 4.8 ± 1.0, 6.1 ± 0.9, 9.4 ± 1.9, 12.9 ± 2.7 and 42 ± 8 µM for L-

817 arginine and 4.5 ± 1.2, 4.8 ± 1.0, 6.4 ± 1.2, 34 ± 6.7, 35 ± 5.1 and 137 ± 26 µM for L-canavanine.

818 # - An inhibition % at 2.5 mM is given in table S2 for compounds which an IC50 was not possible to

819 obtain.

820

821 Table 2: ArgC validated fragment hits.

% DSF ΔTm SPR Binding Binding site Compound compound structure inhibition (°C) Response (RU) (X-ray) at 2 mM H N

NMR322 +3.1 80.5 Substrate 45% O OH

O NH2 NMR401 +4.8 36.7 Co-factor 10% NO OHO

NMR571 +3.1 78.0 Substrate 37%

O O OH NMR863 +5.0 41.6 Co-factor 12% N O O

19.5 (1.25 NADP +0.6 Co-factor - mM) 16.3 (1.25 NADPH -0.9 Co-factor - mM) 822

823

824 Table 3: ArgD validated fragment hits.

SPR Binding DSF ΔTm Response with % inhibition Compound Fragment structure (°C) PLP injection at 2 mM (RU) OH

NMR608 OH 2.4 50.3 16.6

O N

NMR868 OH 0.9 ND ND O O 825

826

827 Table 4: ArgF validated fragment hits.

ITC Compound Fragment structure DSF ΔTm (°C) Kd (µM) S

NH2 NMR007 N 3.0 161 ± 20

O I NMR078 O 4.5 ND

HO OH NMR288 3.5 ND

H F3C N NMR464 4.1 ND N Br CN NMR502 3.5 ND HO F

O2N NMR801 OH 3.1 ND B OH Cl

NMR812 3.6 120 ± 30 OH Cl B OH OH B NO NMR817 HO 2 3.1 ND

L-ornithine 1.5 ND carbamoyl * * phosphate L-citruline 1 ND 828 * - Carbamoyl phosphate is an unstable compound with and half-life time of ≈5 min at 37 °C (42) and

829 therefore a thermal shift and Kd for this molecule could not be determined.

830 ND - not determined. Attempts were made to determine the Kd for these ligands but without success.

831

832 Figure legends

833

834 Figure 1: The L-arginine biosynthesis pathway in Mycobacterium tuberculosis (a). M. tuberculosis L-

835 arginine biosynthesis operon (b).

836

837 Figure 2: (A) X-ray crystal structure of ArgC apoenzyme superposed with the NAPD bound holoenzyme.

838 X-ray crystal structures of ArgC in complex with fragments NMR322 (a) and NMR571 (b) binding to the

839 substrate site and NMR401 (c) and NMR863 (d) binding to the co-factor site. Hydrogen bonds are

840 represented by black dashed lines.

841

842 Figure 3: (a) X-ray crystal structure of M. tuberculosis ArgD showing the dimer. Each protomer of the

843 dimer is highlighted in a different colour. The ArgD protomer comprises of three domains shown in

844 gold (N-terminal domain), green (PLP-binding α/β domain) and cyan (C-terminal domain) (b). X-ray

845 crystal structures of ArgD in complex with fragments NMR608 (c) and NMR868 (d). Hydrogen bonds

846 are represented by black dashed lines.

847

848 Figure 4: (A) X-ray crystal structure of M. tuberculosis ArgF in complex with carbamoyl phosphate.

849 Hydrogen bonds are depicted as black dashed lines. Two of the protomers of the trimer are visible and

850 are coloured differently. (B) Structure of the ArgF trimer bound with fragments at interfacial site.

851 Three X-ray crystal structure of ArgF in complex with different fragments were superposed with the

852 apo structure to create this figure. (C) X-ray crystal structures of ArgF in complex with NMR007 (C),

853 representing the group that binds to subsite 1, NMR801 (D), the single representative of the group

854 that binds to subsite 2 and NMR812 (E), representing the group of fragments that bind to both

855 subsites, with hydrogen bonds depicted as black dashed lines.

856

857 Figure 5: (A) Overlap of X-ray crystal structures of protomers of Apo-ArgB, ArgB co-crystallized with N-

858 acetyl glutamate (NAG), and ArgB co-crystallized with L-arginine. (B) Structure of the ArgB hexamer

859 with fragments bound at the interfacial site of two protomers. Each ArgB protomer is coloured

860 differently. X-ray crystal structures of ArgB in complex with NMR711 (c) and NMR446 (d). Hydrophobic

861 interactions are depicted in green dots, weak hydrogen bonds in orange dots and carbon-π

862 interactions in yellow disks. Only one binding conformation is shown for clarity in both panels. [Fo -

863 Fc] “Omit maps” of NMR711 (e) NMR446 (f) contoured at 1.5σ. These maps were generated with using

864 the phases from the final model. The two adopted conformations are shown for both compounds.

865

866 Figure 6: ArgB inhibition by arginine analogs and allosteric fragment inhibitors. (a) Inhibition of ArgB

867 activity by NMR711, NMR446, L-arginine and L-canavanine. (b) Lineweaver-burk plots for NMR711,

868 NMR446, L-arginine and L-canavanine. Average of replicates and standard deviation are ploted (n=3).

869

870 Figure 7: Dose response curves of inhibitor compounds for inhibition of M. tuberculosis growth

871 (measured as optical density at 600nm) in the presence (a-c) or in the absence (d-f) of 1mM L-arginine.

872 Data is represented as percentage growth of M. tuberculosis strains in the presence of different

873 concentrations of the inhibitor compared to growth in the presence of just vehicle control (no drug).

874 H37Rv (a,d), complemented ∆argB (∆argB-c; b,e), and ∆argB (c,f). Data are representative of one of

875 three independent experiments. Error bars, mean s.d. (n = 3). Compound 1 (green circles), Compound

876 2 (red circles) and L-Canavanine (blue circle).

877

35 bioRxiv preprint doi: https://doi.org/10.1101/2021.03.12.435067; this version posted March 12, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Figure 1 L-glutamate

ArgA (Rv2747) a Acetyl recycled N-acetylglutamate

ArgB (Rv1654) N-acetyl-L-glutamyl-phosphate

ArgJ (Rv1653) ArgC (Rv1652) N-acetyl-L-glutamyl-semialdehyde

ArgD (Rv1655) N-acetylornithine

L-ornithine

ArgF (Rv1656) L-citrulline

ArgG (Rv1658) L-arginosuccinate

ArgH (Rv1659) b L-arginine

Genomic location argC argJ argB argD argF argR argG argH NC_000962.3: NC_000962.3: 1865576 bp 8475 bp 1874051 bp bioRxiv preprint doi: https://doi.org/10.1101/2021.03.12.435067; this version posted March 12, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Figure 2 a b c N320 S186 G187 Y211 S189 Y20 H217 A21

R193 G19

S18 L88 S50 A48 d e f

Y211

H217

T325

R193 bioRxiv preprint doi: https://doi.org/10.1101/2021.03.12.435067; this version posted March 12, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Figure 3 a b C-terminal domain

PLP-binding α/β domain Lysine-Pyridoxal-5’- N-terminal domain Phosphate (lys253) c d

PLP PLP

S281 N81 N81 S281 L82 R142 L82 R142

Y23 Y23 bioRxiv preprint doi: https://doi.org/10.1101/2021.03.12.435067; this version posted March 12, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Figure 4 a b

H128 R101 T53 Q131 Helix 2

T51 C264

R52 Helix 3

c d e T76

R52 E82 V71 V71 E82 R52 T51 I45 R52 R292 R292 S56 F55 L84 F55 T87 A43 L265 D86 Y94 V90 Y94 bioRxiv preprint doi: https://doi.org/10.1101/2021.03.12.435067; this version posted March 12, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Figure 5 a b

Fragment binding site

Interface L-arginine site binding site

NAG binding site e c d

R173 R173 L134 D167 V125 V125 I127 L168 L168 L168 L168 I127 V125

R173 V125 L134 R173 L134 bioRxiv preprint doi: https://doi.org/10.1101/2021.03.12.435067; this version posted March 12, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Figure 6 a

b bioRxiv preprint (which wasnotcertifiedbypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.It

a doi: Figure 7 b c https://doi.org/10.1101/2021.03.12.435067 made availableundera ; CC-BY 4.0Internationallicense this versionpostedMarch12,2021. e d f . The copyrightholderforthispreprint